{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/direction-aware-spatial-context-features-for","title":"Direction-aware Spatial Context Features for Shadow Detection","arxiv_id":"1712.04142","date":"2017-12-12","proceeding":"CVPR 2018 6","authors":["Xiaowei Hu","Lei Zhu","Chi-Wing Fu","Jing Qin","Pheng-Ann Heng"],"abstract":"Shadow detection is a fundamental and challenging task, since it requires an\nunderstanding of global image semantics and there are various backgrounds\naround shadows. This paper presents a novel network for shadow detection by\nanalyzing image context in a direction-aware manner. To achieve this, we first\nformulate the direction-aware attention mechanism in a spatial recurrent neural\nnetwork (RNN) by introducing attention weights when aggregating spatial context\nfeatures in the RNN. By learning these weights through training, we can recover\ndirection-aware spatial context (DSC) for detecting shadows. This design is\ndeveloped into the DSC module and embedded in a CNN to learn DSC features at\ndifferent levels. Moreover, a weighted cross entropy loss is designed to make\nthe training more effective. We employ two common shadow detection benchmark\ndatasets and perform various experiments to evaluate our network. Experimental\nresults show that our network outperforms state-of-the-art methods and achieves\n97% accuracy and 38% reduction on balance error rate.","url_abs":"http://arxiv.org/abs/1712.04142v2","url_pdf":"http://arxiv.org/pdf/1712.04142v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"direction-aware-spatial-context-features-for","repo_url":"https://github.com/stevewongv/dsc-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"direction-aware-spatial-context-features-for","repo_url":"https://github.com/xw-hu/DSC","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"detecting-shadows","task_name":"Detecting Shadows"},{"task_slug":"shadow-detection","task_name":"Shadow Detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/salient-object-detection-on-istd","task":"RGB Salient Object Detection","dataset":"ISTD","model":"DSC","rank_in_archive_order":5,"of":7,"metrics":{"Balanced Error Rate":"8.24"},"uses_additional_data":false},{"leaderboard":"/sota/salient-object-detection-on-sbu","task":"RGB Salient Object Detection","dataset":"SBU / SBU-Refine","model":"DSC","rank_in_archive_order":2,"of":7,"metrics":{"Balanced Error Rate":"5.59"},"uses_additional_data":false},{"leaderboard":"/sota/salient-object-detection-on-ucf","task":"RGB Salient Object Detection","dataset":"UCF","model":"DSC","rank_in_archive_order":4,"of":7,"metrics":{"Balanced Error Rate":"8.10"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.04142","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}